In the field of NBA Draft scouting, projecting jumpshooting development remains an inexact science. As the three-point shot has gained extreme importance over the past decade, teams have placed an increasingly large amount of importance on finding players whose outside-shooting prowess will translate from the college to professional level. However, the traditional jumpshooting indicators (three-point percentage, free-throw percentage, etc.) have proven to be a helpful, yet incomplete projection for shot development in the NBA. Consider the recent example of Tyrese Maxey, taken 21st overall by the Sixers in the 2020 NBA Draft. Taking his college statistics at face value, Maxey projected as a negative shooter at the next level, having possessed a paltry 28% 3-point percentage at Kentucky. That being said, within two seasons, his NBA 3-point percentage has jumped over 10 full percentage points from his college days, and shows no signs of slowing down. On the flipside, the Boston Celtics selected Aaron Nesmith seven picks before Maxey in the same draft. Nesmith was a lights-out shooter at Vanderbilt, pouring in threes at a 52% clip before a foot injury ended his historically good pace. After arriving in Boston, however, Nesmith seems to have lost his once-promising shooting touch, as his percentages dropped precipitously, leading to his removal from the rotation and eventually from the Celtics altogether, as he was recently shipped to Indiana as part of the Malcolm Brogdon trade.
All that being said, percentages alone should not be the primary factor in deciding whether or not a player is an NBA-level shooting prospect. While they are a good starting point, more context is likely needed to present the full analysis of a prospect's shooting profile. As such, in this article I will attempt to evaluate another potential heuristic for jumpshot projection - shot distance, particularly focusing on jumpshots attempted from NBA range during a player's college career.
Over the years, the NCAA has adjusted their three-point line several times, most recently moving it back to match the FIBA international line of 22' 1.75" in 2019. As such, college three-point percentages from before 2019 may be slightly inflated due to the line being closer to the basket. That being said, outside of a brief period in the mid-1990s, the NBA three point line has remained mostly stable throughout its history, lying at 23' 9" along its circumference, and 22' in both corners. This makes outside shooting percentages more comparable across different decades and eras. That being said, since the NBA line is still several feet behind the college line, these so-called "deep-range" threes occur less frequently at the NCAA level. However, because these shots would still have counted for three points in the NBA (unlike closer-in threes, which would have qualified as long twos), I wanted to see if players who shot more frequently or accurately from these distances had a better immediate and overall shooting outlook in the NBA.
Three-Point Line Dimensions at Various Levels (credit: The Seattle Times)
The main dataset I chose for this evaluation came from Will Schreefer, formerly of the now-defunct draft website The Stepien, and can be found at this link. It contains location data for over 1.2 million shots from six NCAA Division-1 seasons, including coordinate data, player information, team matchups, and (most importantly for me) whether or not the shot would have occurred from behind the NBA 3-point line.
For NBA statistics, I used the readily available resources over at sports-reference, who provide tables for every NBA rookie class and both their first-year and career stats. I compiled both the rookie and career statistics for all rookies between the years 2013 and 2019, to match the timeframe of Will's dataset. A sample table, from the 2019 NBA draft, can be found by following this link.
Both of these datasets were merged and queried in various ways depending on the question being asked (more specifics will be provided later).
A sample of the datasets can be found below:
| Year | GameID | ShotID | ID | Period | Shot Full Text | PlayerID (ESPN) | PlayerID (RealGM) | RawX | RawTop | ... | PlayerID | RawLeft | TextDist | FinalRegion | AwayTeam | HomeTeam | AwayID | HomeID | PlayerTeamID | Shot Distance | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 2013 | 400546943 | shot0 | 0.0 | 1.0 | John Puk made Jumper. Assisted by Peter Hooley. | 46282.0 | 17589.0 | 22.0 | 32.0 | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 18.244366 |
| 1 | 2013 | 400546943 | shot1 | 1.0 | 1.0 | Sam Rowley missed Layup. | 56323.0 | 30980.0 | 6.0 | 58.0 | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 4.032381 |
| 2 | 2013 | 400546943 | shot2 | 2.0 | 1.0 | DJ Evans missed Three Point Jumper. | 61855.0 | 43213.0 | 26.0 | 22.0 | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 24.176106 |
| 3 | 2013 | 400546943 | shot3 | 3.0 | 1.0 | DJ Evans made Jumper. | 61855.0 | 43213.0 | 6.0 | 80.0 | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 15.008667 |
| 4 | 2013 | 400546943 | shot6 | 6.0 | 1.0 | Gary Johnson made Layup. | 61856.0 | 43212.0 | 9.0 | 46.0 | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 3.936001 |
5 rows × 34 columns
| Rk | PlayerText | Debut | Age | Yrs | G | MP | FG | FGA | 3P | ... | PF | PTS | FG% | 3P% | FT% | MP.1 | PTS.1 | TRB.1 | AST.1 | -9999 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | Jaylen Adams | Oct 17 '18 ATL @ NYK | 22 | 2 | 41 | 446 | 39 | 118 | 25 | ... | 46 | 110 | 0.331 | 0.329 | 0.778 | 10.9 | 2.7 | 1.5 | 1.6 | adamsja01 |
| 1 | 2 | Deng Adel | Jan 19 '19 CLE @ DEN | 21 | 1 | 19 | 194 | 11 | 36 | 6 | ... | 13 | 32 | 0.306 | 0.261 | 1.000 | 10.2 | 1.7 | 1.0 | 0.3 | adelde01 |
| 2 | 3 | DeVaughn Akoon-Purcell | Oct 23 '18 DEN vs. SAC | 25 | 1 | 7 | 22 | 3 | 10 | 0 | ... | 4 | 7 | 0.300 | 0.000 | 0.500 | 3.1 | 1.0 | 0.6 | 0.9 | akoonde01 |
| 3 | 4 | Rawle Alkins | Dec 17 '18 CHI @ OKC | 21 | 1 | 10 | 120 | 13 | 39 | 3 | ... | 7 | 37 | 0.333 | 0.250 | 0.667 | 12.0 | 3.7 | 2.6 | 1.3 | alkinra01 |
| 4 | 5 | Grayson Allen | Oct 22 '18 UTA vs. MEM | 23 | 4 | 192 | 4198 | 612 | 1412 | 355 | ... | 267 | 1806 | 0.433 | 0.393 | 0.841 | 21.9 | 9.4 | 2.6 | 1.5 | allengr01 |
5 rows × 29 columns
| Rk | PlayerText | Debut | Age | Yrs | G | MP | FG | FGA | 3P | ... | PF | PTS | FG% | 3P% | FT% | MP.1 | PTS.1 | TRB.1 | AST.1 | -9999 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | Steven Adams | Oct 30 '13 OKC @ UTA | 20 | 1 | 81 | 1197 | 93 | 185 | 0 | ... | 203 | 265 | 0.503 | NaN | 0.581 | 14.8 | 3.3 | 4.1 | 0.5 | adamsst01 |
| 1 | 2 | Giannis Antetokounmpo | Oct 30 '13 MIL @ NYK | 19 | 1 | 77 | 1897 | 173 | 418 | 41 | ... | 173 | 525 | 0.414 | 0.347 | 0.683 | 24.6 | 6.8 | 4.4 | 1.9 | antetgi01 |
| 2 | 3 | Pero Antić | Oct 30 '13 ATL @ DAL | 31 | 1 | 50 | 925 | 123 | 294 | 56 | ... | 126 | 352 | 0.418 | 0.327 | 0.758 | 18.5 | 7.0 | 4.2 | 1.2 | anticpe01 |
| 3 | 4 | Chris Babb | Mar 1 '14 BOS vs. IND | 23 | 1 | 14 | 132 | 8 | 30 | 6 | ... | 13 | 22 | 0.267 | 0.222 | NaN | 9.4 | 1.6 | 1.2 | 0.2 | babbch01 |
| 4 | 5 | Anthony Bennett | Oct 30 '13 CLE vs. BRK | 20 | 1 | 52 | 663 | 80 | 225 | 13 | ... | 93 | 217 | 0.356 | 0.245 | 0.638 | 12.8 | 4.2 | 3.0 | 0.3 | bennean01 |
5 rows × 29 columns
For each part of this analysis, I will analyze three aspects of the college data: number of attempts from NBA range in each player's peak college season, percentage made in each player's peak college season, and average distance on all NBA-range threes taken in the player's whole college career. In order to examine if a general trend exists, I will conduct a linear regression analysis on the data. Firstly, we will observe the relationship between the aformentioned three college statistics and players' rookie year three-point percentage.
CRITERIA: Players must have attempted at least 50 NBA-range threes in their peak college season, and at least 50 threes in their rookie year in the NBA.
We can observe a slightly positive trend in the data, but certainly not enough to where we can confirm a trend exists here. That being said, we do notice some outliers on the far right (Trae Young, Justin Jackson, and Carsen Edwards) who may be influencing the slope of this line a bit. Removing these outliers produces the following:
The positive trend here is still pretty weak, although it is orders of magnitude stronger than the model which included the outliers (if you consider a change in slope from 10^-5 to 10^-3 to be significant orders of magnitude). There may be a (very weak) relationship here, but any positive trend at all goes along with what one would expect - more NBA-range threes taken in college should lead to a higher three-point percentage in the Association.
As one might expect, a much stronger positive relationship exists between NBA-Range 3P% and Rookie Year 3P%. I really enjoyed seeing the data backing up this preconceived notion, as it does appear that deep-range shooting accuracy translates quite well to the professional level. The combination of these last two graphs indicates that scouts should reward more selective shooters over those who simply throw up any sort of shot from range. Interestingly, many of the high-volume shooters (Trae Young, Donovan Mitchell, etc.) shot around 36% from NBA range in their peak college season, and around 33% in their rookie season. As such, 33% may be a fair rookie projection for players who are expected to enter the league with a heavy jumpshooting-centric offensive workload.
For this analysis, I used career totals of NBA-Range 3s attempted, rather than solely focusing on the peak season. I chose this route because accuracy is not relevant for the independent variable here, and as such, the distance of threes attempted probably wouldn't change by more than a few inches over a player's whole career. This enabled me to gain a larger sample size for my regression.
Players still needed to have attempted at least 50 NBA-range 3s in college, but over their whole career instead of just a single season
Players near the left side of the graph tended to take more corner threes, as these brought their average distance closer to the 23'9" cutoff line (Jake Layman, Bryn Forbes, and Abdel Nader took so many of them that their average distances were actually less than 23'9"). The trendline here is basically a flat line, indicating that no real relationship exists between "deeper" range and rookie year accuracy. Shooters who tended to select shots from the top of the key or the wings were just as accurate on average coming into the NBA as those who survived on corner shots.
Now that we have seen how each of the three metrics impact jumpshooting as soon as a player enters the league, I wanted to try to see if any of them could predict how a player's shooting prowess will develop over their NBA career. Admittedly, the formulae that I used to calculate shooting development is a bit rough, as I wanted to save myself from having to use a web-scraper and subsequently getting banned from sports-reference's websites. As such, I used the formula (forgive me for the crude notation, LaTeX was not cooperating with me):
(Career 3P% - Rookie 3P%)/(Number of Years in the NBA)
While I do realize this is not a perfect representation of year-by-year development, it does provide a bit of context for how a player's three-point shot has changed over the course of their NBA career, weighted on how long they've played in the NBA.
CRITERIA: Players must have attempted at least 50 NBA-range threes in college, 50 threes in their rookie season, 200 threes in their career, and played at least 3 seasons in the NBA
A slightly positive trend exists here, but again, we have two outliers (Young and Jackson) who may be throwing off this trend slightly. Removing them results in the following:
This time, removing Young and Jackson did little to adjust the slope of the graph, but as we can see, attempting a larger number of NBA-range threes in college does seem to project a (very, very slightly) higher prognosis for jumpshooting development. That being said, the residuals are all over the place here, so there may not be enough to separate this trend from simple randomness.
Outliers exist here (Bryn Forbes and Matt Thomas) so we will remove them and see if it changes the data at all:
The trendline here does appear to be strongly negative, however, much of this can be attributed to sample size. A few points on the left end of the graph (namely Cody Martin, Damion Lee, and Jerian Grant) improved their 3P% more than anyone else in the entire set, likely weighting the slope to be higher in that range. However, one observation that we can make from this graph is the greatly differring collection of player profiles hovering around the 0.45 range - elite sharpshooters like Cameron Johnson, Devin Booker, and Buddy Hield, who came in with high three-point percentages and as such had little room to improve, sit in the same range as noted non-shooting threats Stanley Johnson, Kris Dunn, and Jerome Robinson.
We may have to adjust our original argument from Part 1, to specify that season-long volume does matter in the sense that players must meet the 50-shot cutoff in their peak season in order for shooting accuracy to best predict NBA shot development. However, low volume NBA-range shooters who attempted only 50 such threes in their whole college career have a much more varied outcome, regardless of how accurate they were.
Trae Young is an absurd outlier here, so we should probably remove him before we make any conclusions in our analysis.
Removing Young changed the slope of our trendline by 0.00001, so we really could have just left him in, but overall, we can see that there is a nice overall trend in this data which does show that players who shoot deeper threes in college tend to improve their three-point shooting more as their NBA career progresses. In particular, greater improvements in shooting seem to happen after the 24.4 foot distance mark, as the datapoints begin to have greater separation from the trendline here.
Overall, I think I was able to gain some insight into how attempting NBA-range shots at the college level affects professional outside shooting, both early on and over the course of a player's early career. From Part 1 of my analysis, I observed that, among players who attempted at least 50 threes in their peak college season, more accurate shooters tended to perform better early on in their careers as opposed to those who attempted a higher volume or a longer distance. Unfortunately, I wasn't able to find readily available shot location data for current college players, so I can't identify prospects who meet this criteria for the upcoming college draft class, but it could be useful for those who do have access to such resources.
Likewise, none of the three metrics were particularly important for shooting development, which was disappointing; however, I think shot distance was the most promising. Players who attempted threes from longer than 24.4 feet saw the largest average percentage gains of any prospect in the set (not including Gary Harris). As such, among those who may have disappointed early on in their NBA careers, this college metric may provide some optimism for growth down the line.
There are a few people who were integral to the development of this article that deserve some recognition:
~ Will Schreefer, for creating the only publicly-available dataset that I could find that included shot location data. The "NBA3" variable was an added bonus. Will can be found on Twitter @refersadness
~ Sports-reference, for being an awesome source of baseline basketball stats
~ My dear colleague, Garrett Johnson from The Looney Bin (shameless website plug) for helping me design this idea and listening to my ramblings as I struggled to find a dataset. Garrett can be found on Twitter @halfawaketakes